Anthropic gives METR staff-level access to monitor AI risks

Anthropic CEO Dario Amodei has given the non-profit research group METR "employee-like" access to its systems. This means METR can monitor the company's AI models from the inside. Amodei has made outside evaluation a central part of his plan to address AI safety concerns. The arrangement lets independent experts check AI systems for risks before they reach the public. The move reflects growing pressure on AI companies to open their work to outside review.